S&P's credit outlook warns that $7 trillion in combined hyperscaler capex through 2030 threatens investment-grade ratings, flagging Amazon for a potential AA downgrade and Oracle for possible junk status.
S&P's credit outlook warns that $7 trillion in combined hyperscaler capex through 2030 threatens investment-grade ratings, flagging Amazon for a potential AA downgrade and Oracle for possible junk status.

S&P Global Ratings has issued a rare systemic credit warning on hyperscale cloud infrastructure, projecting that the six largest providers will spend more than $7 trillion combined on capital expenditures through 2030.
"Even the most strongly capitalized, most profitable companies could suffer significant pain in an economic downturn," S&P analysts wrote in the report, which noted that none of the six hyperscalers have quantitatively disclosed AI investment returns, forcing the rating agency to rely on benchmark assumptions that current and future spending will ultimately generate returns.
The credit outlook flags divergent risk profiles across the sector. Amazon faces potential AA rating pressure around 2027 unless earnings improve or capital spending moderates, S&P said. Oracle, rated BBB-, could fall into high-yield territory on a downgrade — a significant market event given its roughly $117 billion in index-eligible dollar bonds. Microsoft's debt, lease and purchase obligations can reach only about one year of EBITDA before triggering rating concerns, while Google's AA+ rating provides more financial headroom.
The warning marks the first time a major credit rating agency has formally assessed the systemic risk of AI infrastructure spending from a debt perspective. Nvidia chief financial officer Colette Kress said on the company's fiscal second-quarter earnings call that top-five hyperscaler capex is expected to reach nearly $800 billion in 2026 and $1.3 trillion in 2027, with cloud industry backlog now exceeding $2 trillion. Nearly half of US private sector economic growth is now tied to AI-related capital spending.
Circular financing links hyperscaler balance sheets
S&P's report specifically calls out the interconnected financing structures that have emerged across the hyperscaler sector. These include residual value guarantees, take-or-pay agreements, chip financing, lease liabilities, power purchase agreements, backstop guarantees and direct equity investments.
"Interconnected financing structures could magnify volatility if demand unexpectedly weakens," the report said. "A default by any one party would ripple through lenders, asset-value guarantors and counterparties awaiting payment."
The concern is that these risks, while manageable at the individual company level, could become highly correlated in a sharp demand downturn. S&P categorized the potential risks across known and unknown, quantifiable and unquantifiable dimensions — including overbuilding risk, circular financing risk, and the potential for open-source models to disrupt commercial AI business models.
Credit headroom splits the six hyperscalers
The rating agency's analysis shows significant divergence in how much financial pressure each hyperscaler can absorb before triggering a downgrade. Microsoft's tolerance for debt, leases and similar obligations is roughly equivalent to one year of EBITDA. Oracle can push adjusted debt leverage to 4.5 times EBITDA before hitting its downgrade threshold. Amazon's path from "hoping investments work" to "investments have worked" appears narrow without either earnings improvement or capex moderation.
For Oracle specifically, the stakes are unusually high. A downgrade from BBB- would push the company's bonds into high-yield territory, triggering forced selling by investment-grade bond funds and index rebalancing across its $117 billion bond footprint. That would raise borrowing costs precisely when the company needs capital for continued AI infrastructure expansion.
The report's baseline scenario assumes all current and future AI investments ultimately convert into EBITDA growth. If that assumption fails, the consequences would extend well beyond individual credit ratings. With Nvidia projecting that hyperscalers represent roughly 24 percent of its data center revenue — about $195 billion at current run rates — the credit health of cloud providers is now a direct input into semiconductor demand forecasts.
Amazon shares and Oracle bonds have not yet fully priced in the scope of S&P's warning. The rating agency noted that rating decisions are made by committee and special circumstances could provide relief, but the structural trajectory remains: $7 trillion in spending, unquantified returns, and circular financing that could turn a demand shock into a credit event.
This article is for informational purposes only and does not constitute investment advice.